Bioelectrical impedance analysis to assess body composition in obese adult women: The effect of ethnicity

Transcription

1 International Journal of Obesity (1998) 22, 243±249 ß 1998 Stockton Press All rights reserved 0307±0565/98 $12.00 Bioelectrical impedance analysis to assess body composition in obese adult women: The effect of ethnicity JM Jakicic, RR Wing and W Lang University of Pittsburgh School of Medicine, Pittsburgh, USA OBJECTIVE: To examine whether the accuracy of bioelectrical impedance analysis (BIA) to estimate body composition in overweight women is affected by the ethnicity of the individuals. DESIGN: Cross-sectional design to compare body composition estimated by BIA to body composition measured by dual energy x-ray absorptiometry (DEXA), which was the reference method. SUBJECTS: One hundred twenty three overweight women participated in this study, of which 43 women were African-American (aged y; BMI, kg/m 2 ) and 80 were Caucasian (aged y; BMI, kg/m 2 ). MEASUREMENTS: Body composition was estimated from BIA using both a generalized and an obesity-speci c equation. These estimations were compared to body composition measured by DEXA, which was the reference method. RESULTS: The generalized BIA equation underestimated lean body mass (LBM) by kg in Caucasian women and kg in African-American women, with the difference between the ethnic groups being signi cant (P < 0.001). The obesity-speci c equation underestimated LBM in Caucasians by kg and overestimated LBM in African-Americans by kg (P < 0.001). An ethnic-speci c equation is proposed, and cross-validation of this equation indicates that it provides a reasonable estimate of body composition in overweight women. CONCLUSIONS: The accuracy of BIA to estimate body composition appears to be affected by the ethnicity of the individual. Therefore, an ethnic-speci c equation for overweight women is proposed. However, further validation of this prediction model in an ethnically diverse population is necessary. Keywords: overweight; lean body mass; percent body fat; dual-energy x-ray absorptiometry; African-American Introduction Correspondence: John M Jakicic, University of Pittsburgh School of Medicine, 3811 O'Hara Street, Pittsburgh, PA 15213, USA. Received 14 February 1997; revised 26 June 1997; accepted 15 October 1997 Bioelectrical impedance analysis (BIA) is an appealing method of estimating body composition because it is simple, quick and noninvasive, compared to other techniques that require expensive equipment (for example, dual energy x-ray absorptiometry (DEXA)), skilled laboratory personnel (for example, anthropometrics, hydrostatic weighing), or a very cooperative participant (that is, hydrostatic weighing). Signi cant correlations have been reported between percent body fat assessed from BIA and both hydrostatic weighing 1,2 and DEXA. 3,4 The standard error for estimating (SEE) percent body fat using empirically derived equations provided by the manufacturer of the BIA device has been shown to be approximately 6%. 1,2 However, the use of age, gender and fatnessspeci c equations may enhance the accuracy of the BIA method to estimate body composition. 5,6 For example, using population-speci c equations, Rising et al 2 have reported a SEE of 3.2% for estimating percent body fat in Pima Indians. Furthermore, Stolarczyk et al 7 have reported that the SEE decreased from 8.1 kg to 2.6 kg when a population-speci c equation was used to estimate lean body mass in Native American women. Despite the development of age, gender, fatness and population-speci c equations (for example, Native Americans, Pima Indians), the majority of studies that have examined the accuracy of BIA have used Caucasians. 8 Kuczmarski 9 has indicated that the lack of valid and reliable ethnic-speci c BIA predictionequations limits the usefulness of BIA data collected in studies that include an ethnically diverse sample (for example, National Health and Nutrition Examination Survey). For example, few studies have examined whether BIA can accurately estimate body composition in African-Americans. Since there are differences in bone mineral density between Caucasians and African-Americans, 10 these differences may have a signi cant impact on the measurement of lean body mass (LBM) using BIA. Consequently, it may be appropriate to examine the accuracy of BIA compared

2 244 to DEXA as the reference method, because of the ability of DEXA to measure both LBM and bone mineral content (BMC). Therefore, the purpose of this study was to examine the effect of ethnicity (Caucasians vs African-Americans) on the accuracy of BIA in estimating body composition in obese adult females, using DEXA as the reference method. Methods Subjects One-hundred twenty three overweight women (43 African-American and 80 Caucasian) participated in this study. All subjects were aged 25±45 y and at least 20 pounds over their ideal body weight according to the Metropolitan Life weight tables. 11 Subject characteristics are presented in Table 1. Subjects were not taking prescription medication that could affect body weight (for example, synthroid) or uid balance (for example, diuretics), nor did they have any metabolic condition that could affect the results of this study. In addition, subjects were excluded if they were currently pregnant or had been pregnant within the previous six months. All subjects gave written informed consent and all procedures were approved by the Institutional Review Board at the University of Pittsburgh prior to beginning this study. Assessments Weight and height. Weight was measured to the nearest 0.25 pounds using a calibrated medical balance-beam scale (Health-O-Meter, Bridgeview, IL), with subjects wearing a light-weight hospital gown and no shoes. Height was measured to the nearest 0.1 cm using a calibrated stadiometer (Perspective Enterprises, Kalamazoo, MI). DEXA. A Lunar DPX-L densiometer (Madison, WI) was used to assess body composition in all subjects, and this was considered the reference method in this study. DEXA provides an assessment of fat mass, LBM, percent body fat and BMC. LBM can be computed to include BMC (LBM total ) or not to include BMC (LBM tissue ). Subjects were clothed in a light-weight hospital gown and all jewelry was removed prior to the DEXA scan being performed. A urine pregnancy test (hcg, Abbott Laboratories, Abbott Park, IL) was performed immediately prior to the scan. Women with a positive pregnancy test were excluded from this study. Each total body scan was performed at the scanning mode recommended by the manufacturer, with radiation exposure ranging from 0.02±0.06 mrem depending on the scanning mode used. Scanning mode was determined according to the thickness of the trunk region, with a thickness of < 22 cm, 22±28 cm or > 28 cm corresponding to a fast (10 min), medium (20 min) or slow (37 min) scanning mode, respectively. The scans were analyzed using Lunar DPX-L software version 1.3. Test-retest using DEXA to assess body composition in our laboratory has shown a difference (mean standard deviation) of % for percent body fat, kg for fat mass, kg for LBM and kg for BMC. For these test-retest measurements, scans were performed on the same day, with the subject removed from the DEXA and repositioned between tests. BIA. Measurement of LBM using BIA was performed on the same morning as DEXA. A RJL BIA- 101A (RJL Systems, Inc., Clinton Twp, MI) four terminal impedance analyzer was used to measure resistance and reactance to a maximum constant current of 500 microamps at a frequency of 50 khz. This instrument was calibrated throughout the study using a 500-O resistor according to the procedures recommended by the manufacturer. Subjects were instructed to fast (except for normal water intake) and abstain from vigorous exercise for 12 h prior to this evaluation. BIA was assessed with subjects in a supine position. The skin surface was cleaned using rubbing alcohol prior to applying Unitrac disposable electrodes. Once the alcohol on the skin surface dried, electrodes were placed on the right side of the body at Table 1 Descriptive characteristics of subjects (mean standard deviation) Total (n ˆ123) Caucasians (n ˆ 80) African-Americans (n ˆ 43) Age (y) Height (cm) Weight (kg) BMI (kg/m 2 ) Resistance Reactance DEXA measurements Percent Body Fat (%) Fat Mass (kg) LBM (kg) Tissue * * BMC (kg) * * * Indicates signi cant difference between Caucasians and African-Americans at P < BMI ˆ body mass index; DEXA ˆ dual energy x-ray absorptiometry; LBM ˆ lean body mass; BMC ˆ bone mineral content.

3 the following four sites: between the styloid processes of the ulna and radius (E1), distal end of the second and third metacarpals (E2), between the lateral malleolus and medial malleolus (E3), and distal end of the rst and second metatarsals (E4). There was a minimum of 8 cm between electrodes E1 and E2, and electrodes E3 and E4. LBM was estimated using a generalized equation 1 and an obesity-speci c equation, 6 and percent body fat was then computed using the following equation: Percent body fat ˆ body weight LBM =body weight 100: Because the subjects being examined in this study were obese, we were interested in using prediction equations that have been validated on obese populations. The equation proposed by Segal et al 1 was derived from a sample of subjects that included both lean and obese individuals, and the study by Fuller et al 12 indicates that this equation provides a more accurate estimate of LBM compared to other existing equations derived on similar samples. Furthermore, Fuller et al 12 showed that the obesity-speci c equation proposed by Segal et al 6 provided a more accurate estimate of LBM in an obese population when compared to other equations derived from primarily obese populations. Statistical analysis Statistical analysis was performed using Statistical Package for the Social Sciences software (SPSS for Windows, version 7.5) 13 and SAS (version 6.1.1). 14 A two-factor (Measurement6Ethnicity) analysis of variance (ANOVA) was used to assess the difference between DEXA and BIA. Multiple regression was performed separately for LBM tissue (tissue of LBM only) and LBM total (tissue of LBM BMC) as the dependent, with weight, height, age and ethnicity entered into each analysis and either resistance (R) or height 2 /R added separately to the analysis (since both R and height 2 /R have been used in previous studies). These regression equations were cross-validated using sub-samples of the original 123 subjects. Statistical signi cance was de ned at P < Results Results of the comparison between African-American and Caucasian subjects are shown in Table 1. There were no signi cant differencse between the ethnic groups for age, height, weight, body mass index (BMI), resistance or reactance. In addition, body composition measured by DEXA showed no difference between the groups for percent body fat or fat mass. However, both components of LBM total, BMC and LBM tissue, were signi cantly greater in African- Americans vs Caucasians (P < 0.001). A two-factor (Measurement6Ethnicity) repeated measures analysis of variance was used to assess differences between DEXA and BIA. The signi cant Measurement6Ethnicity interaction (P < 0.001) indicates that the difference in LBM measured by DEXA (LBM total ) vs BIA using the generalized equation, 1 was greater in Caucasians compared to African-Americans (see Table 2). Results showed that BIA overestimated LBM in Caucasians by kg vs kg in African-Americans (P < 0.001). As a result, percent body fat using BIA was underestimated by % in Caucasians vs % in African- Americans, and the Measurement6Ethnicity interaction was statistically signi cant (P < 0.004). Data were also analyzed to examine the accuracy of an obesity-speci c equation, 6 to estimate LBM using BIA (see Table 2). Again, a signi cant (P < 0.001) Measurement6Ethnicity interaction was found, with BIA overestimating LBM by kg in Caucasians and underestimating LBM by kg in African-Americans. This translated into percent body fat being underestimated by % in Caucasians and overestimated by % in African- Americans, with the Measurement6Ethnicity interaction again being statistically signi cant (P 0.001). Results for each of the regression analyses to predict LBM total using height 2 /R and R are presented in Table 3. As shown in the table, regardless of whether height 2 /R or R were used in the equation, the demographic s of height, weight, and ethnicity were all signi cant contributors to the prediction of LBM total (P < 0.05). These data suggest that including ethnicity in the BIA equations to estimate 245 Table 2 Repeated measures analysis of variance to compare body composition by ethnic group using dual energy x-ray absorptiometry (DEXA) and bioelectrical impedance analysis (BIA), using generalized and obesity-speci c equations Body composition measures P-values for DEXA vs BIA a P-values for DEXA vs BIA b DEXA BIA a BIA b Measure Ethnicity Measure 6 Ethnicity Measure Ethnicity Measure 6 Ethnicity Lean body mass (kg) Caucasians (n ˆ 80) < < < African-Americans (n ˆ 43) Percent Body Fat Caucasians (n ˆ 80) < African-Americans (n ˆ 43) a using generalized equation from Segal et al (1985). 1 b using obesity-speci c equation from Segal et al (1988). 6

4 246 Table 3 Multiple regression to predict tissue of lean body mass bone mineral content (LBM total ) using bioelectrical impedance analysis (BIA) and descriptive characteristics of subjects Dependent (using DEXA) Predictor a Coefficient P-values for each independent Predicted residual sum of squares R 2 Mean square error Equation 1 LBM total (kg) Intercept 2.04 Weight (kg) 0.19 < Resistance (R) < Ethnicity 2.63 < Height (m) < Equation 2 LBM total (kg) Intercept 2.68 Weight (kg) 0.19 < height 2 /R 0.20 < Ethnicity 2.55 < Height (m) a predictor s presented in the order in which they entered the model. Ethnicity: 0 ˆ Caucasian, 1 ˆ African-American. height 2 /R ˆ height (cm) 2 /Resistance. LBM will account for the 2.5 kg difference in LBM shown between the overweight Caucasian and African-American women in this study (see Table 1). Based on the data used in this study, examination of the residual scores showed no signi cant outliers. The data were examined further to assess whether there were differences in the accuracy of the prediction of LBM or the t of the model, when either height 2 /R or R was used in the equation. The predicted residual sum of squares, r 2, and the mean square error were similar for both equations, indicating no difference between the two equations in the accuracy of the prediction or the t of the model (see Table 3). Results of regression analysis to predict LBM tissue were virtually identical to those described for LBM total (data not shown). We applied these equations to the sample of 123 subjects in this study, and compared the LBM total to the LBM predicted from these equations. These results indicate that the equation which uses height 2 /R may provide a more accurate estimate of LBM, and these results are presented in Table 4. Based on these results, the remaining analyses were performed on the equations that used LBM total and height 2 /R. Random samples of 30 (21 Caucasian, 9 African- American) and 60 (39 Caucasian, 21 African-American) subjects were selected from the original 123 subjects for the purpose of cross-validation of the regression equations shown in Table 3. Using the regression equation that included height 2 /R, results showed no signi cant difference between the observed and predicted LBM total, with these values being kg and kg, respectively. Similar results were found when the equation was applied to the random sample of 60 subjects, with the observed LBM total being kg and the predicted LBM total being kg (NS). In addition, there was no signi cant difference between the observed and the predicted LBM total when analyzed separately for Caucasians and African-Americans in either of the random samples of 30 or 60 subjects. To examine further the application of an ethnicity speci c equation to estimate body composition, we randomly divided the original 123 subjects into two Table 4 Comparison by ethnicity of lean body mass (LBM) measured by dual energy x-ray absorptiometry (DEXA) and predicted from the proposed bioelectrical impedance analysis (BIA), ethnic-speci c equations for 123 overweight women (80 Caucasian and 43 African- American) DEXA (mean s.d.) BIA (mean s.d.) Mean difference LBM total (kg) a Caucasians (n ˆ 80) * African-Americans (n ˆ 43) * LBM total (kg) b Caucasians (n ˆ 80) African-Americans (n ˆ 43) a BIA computed using equation from Table 3 which incorporates resistance in the model. b BIA computed using equation from Table 3 which incorporates height 2 /R. * indicates statistically signi cant at P < LBM total ˆ tissue of lean body mass bone mineral content.

5 Table 5 Multiple regression to predict tissue of lean body mass bone mineral content (LBM total ) using bioelectrical impedance analysis (BIA) and descriptive characteristics in sub-samples of 30 and 60 subjects 247 Sample size Dependent (using DEXA) Predictor Coefficient P-values for each independent Predicted residual sum of squares R 2 Mean square error 30 a LBM total (kg) Intercept Weight (kg) ht 2 /R Ethnicity Height (m) b LBM total (kg) Intercept 5.49 Weight (kg) 0.17 < ht 2 /R 0.29 < Ethnicity Height (m) Ethnicity: 0 ˆ Caucasian, 1 ˆ African-American. ht 2 /R ˆ height (cm) 2 /Resistance. a indicates 21 Caucasians and 9 African-Americans. b indicates 39 Caucasians and 21 African-Americans. groups which consisted of 60 (39 Caucasian, 21 African-American) and 63 subjects (41 Caucasian, 22 African-American). We generated regression models to estimate LBM total for each of these groups (see Table 5) and found that these equations were similar to the original equation shown in Table 3. Furthermore, when we applied the regression equation generated on the sample of 60 to the sample of 63, no signi cant difference was shown between the observed ( kg) and predicted ( kg) LBM total, and the results were similar when separated by ethnicity. Discussion The results of this study support previous studies suggesting that BIA overestimates LBM in obese adults when a generalized equation 1 is used to predict LBM. Valero et al 15 examined the accuracy of BIA in 10 obese patients and showed LBM to be overestimated by approximately 7 kg when compared to DEXA, and a similar pattern was reported by Beshyah et al 16 in a sample of obese hypopituitary patients. However, when separated by ethnicity, the results of this study suggest that the error in the estimate of BIA using a generalized equation is primarily a result of the error in Caucasian subjects not in African-Americans (see Table 2). The ethnicity of the sample population used to derive the prediction equation proposed by Segal et al 1 was not reported, but the majority of studies of BIA have been undertaken using, primarily, a Caucasian population. 8 Because these prediction equations are based on data from, primarily, Caucasian populations, one might expect these equations to be more accurate for Caucasians than African-Americans. However, the result of this study do not con rm this hypothesis, and are contrary to what was expected. To improve the estimate of body composition using BIA in obese adults, Segal et al 6 have proposed an obesity-speci c equation. The results of the current study show that the accuracy of the estimate of body composition using BIA for obese Caucasian women is improved when an obesity-speci c prediction equation is used, with the overestimation of LBM reduced from 2.6 kg to 0.9 kg (Table 2). These ndings are very similar to the results reported by Fuller et al 12 in a study of 15 obese Caucasian women. In this previous study, the generalized equation 1 overestimated LBM by kg and the obesity-speci c equation 6 reduced this to kg. In contrast, for the obese African-American women, the generalized equation 1 provided a reasonable estimation of LBM (see Table 2), whereas use of the obesity-speci c equation resulted in BIA underestimating LBM by 1.2 kg. Therefore, the BIA equation used to predict LBM for obese African-American women may need to be different from the equation used for obese Caucasian women. The results of this study suggest that the ethnicity of the subjects can affect the magnitude and the direction of the error in the estimation of body composition when using BIA. This may be due to African-Americans having greater body density 17 and greater BMC 10 compared to Caucasians, and these differences have not been accounted for in the existing BIA equations. Therefore, we propose the use of an ethnic-speci c BIA equation to estimate body composition (Table 3, Equation 2). A search of MEDLINE (National Library of Medicine, Bethesda, MD) databases from 1966±1996 was performed using BIA and African-Americans as keywords. When combined, these keywords indicated that there were only four published articles, with none of these articles published prior to Furthermore, only one of these articles examined the accuracy of BIA in African-Americans, supporting the belief that research in this area is lacking. 8,9 Clark et al 18

6 248 reported that BIA overestimated percent body fat by 6% in African-American collegiate football players, when hydrostatic weighing was used as the criterion measure. In comparison, BIA overestimated percent body fat by 3% in Caucasian athletes in the same study. We were unable to nd any studies that examined the accuracy of BIA in obese African- American women. Given this, researchers and clinicians should consider using the ethnic-speci c equations proposed in this study when assessing body composition in a sample of obese women that include both Caucasians and African-Americans. These equations accounted for the 2.5 kg difference in LBM between Caucasians and African-Americans who participated in this study (Table 1, Table 3). Results of cross-validation analyses showed that the proposed BIA equation using height 2 /R (Table 3, Equation 2) provides an accurate estimate of LBM in overweight Caucasian and African-American women. Despite these promising ndings, the accuracy of this equation needs to be further validated in other samples which include both Caucasians and African-Americans. The need for an ethnic-speci c equation to enhance the prediction of body composition using BIA may suggest that there is biological and/or metabolic difference between Caucasians and African-Americans. Schutte et al 17 have suggested that the biological and physiological assumptions related to the measurement of LBM which are based largely on Caucasian samples may not be accurate for African-Americans. For example, it has been suggested that the density of muscle and bone may be greater in African-Americans compared to Caucasians. 10,19 Thus, because the majority of BIA prediction equations are based on Caucasian samples, the impact that this difference in the density of LBM has on the accuracy of BIA has not been thoroughly examined. The accuracy of both DEXA and BIA may be affected by altered uid distribution and volume. 20 Because an overweight population was examined in this study, it is possible that edema may have been present which could have affected the accuracy of both the DEXA and BIA. We did not conduct a clinical examination to assess whether this condition was present in any of the subjects in this study. However, based on self-reported medical histories, none of the subjects in this study reported edema, nor were any of the subjects taking medication that would affect uid balance (for example, diuretics). However, future studies examining the accuracy of BIA, that use DEXA as the reference method, should consider examining this parameter and the impact this may have on body composition in both Caucasian and African-American subjects. In summary, few studies have examined the accuracy of BIA to estimate body composition in speci c ethnic groups (for example, African-Americans). The results of this study suggest that the use of a commonly cited generalized BIA prediction model 1 can accurately estimate LBM in overweight African- American women, but this same equation overestimates LBM in overweight Caucasian women. The use of an obesity-speci c equation 6 improved the estimation of LBM using BIA for the Caucasian women in this study, but decreased the accuracy for the African- American women. Therefore we propose the use of a prediction model that accounts for potential differences between these two ethnic groups (see Table 3), and the results of this study indicate that these equations may be expected to perform reliably for African-American and Caucasian women with characteristics similar to the subjects used in this study. Therefore, the ndings of this study are limited to overweight Caucasian and African-American women; thus future studies should examine whether the use of an ethnicity-speci c BIA equation is also applicable in men and leaner populations. Acknowledgments This study was supported by the National Institutes of Health with awards to Dr Wing (DK29757) and Dr Jakicic (HL561227), and the Obesity/Nutrition Research Center at the University of Pittsburgh (DK46204). References 1 Segal KR, Gutin B, Presta E, Wang J, Van Itallie TB. Estimation of human body composition by electrical impedance methods: a comparative study. J Appl Physiol 1985; 58: 1565± Rising R, Swinburn B, Larson K, Ravussin E. Body composition in Pima Indians: validation of bioelectrical impedance. Am J Clin Nutr 1991; 53: 594± Stewart SP, Bramley PN, Heighton R, Green JH, Horsman A, Losowsky MS, Smith MA. Estimation of body composition from bioelectrical impedance of body segments: comparison with dual-energy x-ray absorptiometry. Brit J Nutr 1993; 69: 645± Brodowicz GR, Mans eld RA, McClung MR, Althoff SA. Measurement of body composition in the elderly: dual energy x-ray absorptiometry, underwater weighing, bioelectrical impedance analysis, and anthropometry. Gerontology 1994; 40: 332± Lukaski HC, Johnson PE, Bolonchuk WW, Lykken GI. Assessment of fat-free mass using bioelectrical impedance measurements of the human body. Am J Clin Nutr 1985; 41: 810± Segal KR, Van Loan M, Fitzgerald PI, Hodgdon JA, Van Itallie TB. Lean body mass estimation by bioelectrical impedance analysis: a four-site cross-validation study. Am J Clin Nutr 1988; 47: 7±14. 7 Stolarczyk LM, Heyward VH, Hicks VL, Baumgartner RN. Predictive accuracy of bioelectrical impedance in estimating body composition in Native American Women. Am J Clin Nutr 1994; 59: 964± NIH Consensus statement. Bioelectrical impedance analysis in body composition measurement. National Institutes of Health Technology Assessment Statement. December 12±14, Nutrition 1996; 12(11±12): 749± Kuczmarski RJ. Bioelectrical impedance analysis measurements as part of a national nutrition survey. Am J Clin Nutr 1996; 64(suppl): 453S±458S. 10 Liel Y, Edwards J, Shary J, Spicer KM, Gordon L, Bell NH. The effects of race and body habitus on bone mineral density

7 of the radius, hip, and spine in premenopausal women. J Clin Endocrinol Metab 1988; 66: 1247± Metropolitan Life Insurance Company. Nutrition and Athletic Performance: Values for Ideal Weight. Metropolitan Life Insurance: New York, Fuller NJ, Sawyer MB, Elia M. Comparative evaluation of body composition methods and predictions, and calculation of density and hydration fraction of fat-free mass, in obese women. Int J Obes 1994; 18: 503± SPSS User's Guide. SPSS, Inc: Chicago, SAS Institute. SAS/STAT User's Guide, Version 6, 4th Edition, SAS Institute: Cary, Valero MA, Leon-Sanz M, Gomez I, Martinez G, Hawkins F.A comparison between double-photon absorptiometry (DEXA), impedance and anthropometry in the study of the body composition of obese subjects. Nutricion Hospitalaria 1994; 9: 12± Beshyah SA, Freemantle C, Thomas E, Page B, Murphy M, Johnson DG. Comparison of measurements of body composition by total body potassium, bioimpedance analysis, and dualenergy x-ray absorptiometry in hypopituitary adults before and during growth treatment. Am J Clin Nutr 1995; 61: 1186± Schutte JE, Townsend EJ, Hugg J, Shoup RF, Malina RM, Blomqvist CG. Density of lean body mass is greater in blacks than in whites. J Appl Physiol 1984; 56: 1647± Clark RR, Kuta JM, Sullivan JC. Cross-validation of methods to predict body fat in African-American and Caucasian collegiate football players. Res Q Exerc Sport 1994; 65: 21± Hampton MC, Huenemann RL, Schapiro LR, Mitchell BW, Behnke AR. A longitudinal study of gross body composition and body conformation and their association with food and activity in a teen-age population. Am J Clin Nutr 1966; 19: 422± Roubenoff R, Kehayias JJ, Dawson-Hughes B, Heyms eld SB. Use of dual-energy x-ray absorptiometry in body composition studies: not yet a `gold standard'. Am J Clin Nutr 1993; 58: 589±